Conventional data compression methods may allow more efficient use of storage media by using algorithms to compress an original data item having a given length measured in bits, into a compressed representation of the data item, which may include less than the original number of bits. Some data patterns, however, may not be well-suited for compression, and if compression is performed blindly on such data patterns, data expansion may occur. In these cases, the original uncompressed data must be re-fetched and written in lieu of the expanded data, which may cause loss of performance and power.